Design and precision construction of novel magnetic

Precision Engineering 31 (2007) 337–350
Design and precision construction of novel magnetic-levitation-based
multi-axis nanoscale positioning systems
Won-jong Kim ∗ , Shobhit Verma, Huzefa Shakir
Department of Mechanical Engineering, Texas A&M University, College Station TX 77843, USA
Received 20 July 2006; received in revised form 30 January 2007; accepted 13 February 2007
Available online 12 April 2007
Abstract
This paper presents two novel six-axis magnetic-levitation (maglev) stages capable of nanoscale positioning. These stages have very simple and
compact structures, which is advantageous to meet the demanding positioning requirements of next-generation nanomanipulation and nanomanufacturing. Six-axis motion generation is accomplished by the minimum number of actuators and sensors. The first-generation maglev stage, namely
the -stage, is capable of generating translation of 300 ␮m and demonstrates position resolution better than 2 nm root-mean-square (rms). The
second-generation maglev stage, namely the Y-stage, is capable of positioning at a resolution better than 3 nm rms over a planar travel range of
5 mm × 5 mm. A novel actuation scheme was developed for the compact structure of this stage that enables six-axis force generation with just three
permanent-magnet pieces. This paper focuses on the design and precision construction of the actuator units, the moving platens, and the stationary
base plates. The performance of the two precision positioners is compared in terms of their positioning and load-carrying capabilities and ease
of manufacture. Control system design for the two positioners is discussed and an experimental plant transfer function model is presented for the
Y-stage. The superiority of the developed instruments is also demonstrated over other prevailing precision positioning systems in terms of the travel
range, resolution, and dynamic range. The potential applications of the maglev positioners include semiconductor manufacturing, microfabrication
and assembly, nanoscale profiling, and nanoindentation.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Precision assembly; Magnetic levitation; Nanopositioning; Multi-DOF motion control
1. Introduction
In the modern era of nanosystems, the precise manufacture of nanoscale parts is a very crucial task [1]. Precision
positioning and vibration isolation play a crucial role while
manufacturing, manipulating, or scanning on the micro/nano
level. Nanopositioning stages need to meet the demanding
positioning requirements of the current and future nanotechnology. These positioning stages along with cutting-edge
manufacturing technologies will lead to the development of
micro-assemblies and nanostructures. The applications of these
stages include microstereolithography (␮STL), nanoscale profiling, and nanoindentation. The key role of a nanopositioning
stage is to load, position, and orient an object and keep it stable
without much vibration or noise. The stage must be able to travel
∗
Corresponding author. Tel.: +1 979 845 3645; fax: +1 979 862 3989.
E-mail address: [email protected] (W.-j. Kim).
0141-6359/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.precisioneng.2007.02.001
in all directions and place the object at the desired position with
minimum error. This requires high accuracy, large travel range,
and simultaneous generation of multi-DOF (degree-of-freedom)
motions and high control bandwidth. However, there are many
theoretical and practical challenges in this early phase of nanopositioning and nanomanipulation research. Important technical
issues include the development of high-precision manipulation
methodologies and the tools for specific applications [2].
Prevailing technologies for ultra-high precision positioning
include scanning probe microscopy (SPM). SPM measures and
images the object’s surfaces on a fine scale, down to the level
of molecules and groups of atoms. It is based on the concept of
scanning an extremely sharp tip (with a 3–50 nm radius of curvature) across the object surface. The tip is mounted on a flexible
cantilever, allowing the tip to follow the surface profile. When
the tip moves in proximity to the investigated object, interaction
forces between the tip and the surface influence the movement
of the cantilever. These movements are detected by selective
sensors. By following a raster pattern, the sensor data forms
338
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
an image of the probe-surface interaction. Three most common
SPM techniques are atomic force microscopy (AFM), scanning
tunneling microscopy (STM), and near-field scanning optical
microscopy (NSOM). Currently most of the atomic-level positioning and profiling is accomplished by AFMs and STMs [3,4].
As an example, the Dimension 3100 SPM of Veeco Instruments
utilizes automated AFM and STM techniques to measure surface
characteristics for semiconductor wafers, lithography masks,
magnetic media, biomaterials, etc. [5]. Although their primary
application was topographical imaging, these instruments have
become the prime tools for nanomanipulation.
Most of the actuation units in micro/nanopositioning devices
are based on piezoelectric materials. Piezoelectric actuators have
become a standard solution in precision positioning applications,
where the travel range must be small. Zhang and Zhu developed
a high-stiffness linear piezoelectric motor with a resolution of
5 nm [6]. This motor is capable of generating a force of 15 N over
a travel range of 5 mm with a maximum step size of 3 ␮m and
various motion speeds with precise positioning down to 0.1 ␮m.
Dong et al. presented a 10-DOF nanorobotic manipulator actuated by lead zirconium titanate (PZT) with a linear resolution
of 30 nm, rotary resolution of 2 mrad, and speed on the order
of several millimeters per second [7]. Among the commercial
semiconductor manufacturing equipment companies, Philips
and ASML have done significant work with maglev short-stoke
stages for lithography. Carter et al. developed a method, apparatus, and system for controlling an electromagnetically levitated
reticle-masking blade in a photolithography system [8]. Hollis and Musits suggested mounting a position sensor equipped
with planar armature, in operative juxtaposition to a stator, by a
spring colonnade basket of flexure columns, which is capable of
providing a low-friction precision x-y-θ positioner [9]. Its operations are computer-controlled with feedback of actual position
for closed-loop servo operation. Galburt invented an electromagnetic alignment apparatus particularly adapted for use in
aligning the wafers in a microlithography system [10].
There are other techniques that have been applied successfully for micromanipulation. Sun et al. fabricated and
experimented on a dual-axis electrostatic microactuation system
[11]. This actuation system has a 0.01 ␮m resolution over a planar motion of 5 ␮m in the x- and y-axes. A high-bandwidth linear
actuator consisting of an air-bearing and voice-coil motor was
fabricated and tested by Mori et al. [12]. This actuator is capable
of 1 nm steps without overshoots or undershoots. Egashira et al.
developed a precision stage using a non-resonant-type ultrasonic
motor with a resolution of 0.69 nm [13]. Culpepper and Anderson came up with an idea of a low-cost nanomanipulator, which
uses a six-axis compliant mechanism driven by electromagnetic
actuators [14]. The stage mechanism is monolithic, and its errors
are less than 0.2% of the full scale for a cubical volume travel
of 100 ␮m. Zyvex Corporation developed a nanomanipulator
that actuates using coarse and fine actuators with 12 mm travel
ranges in x, y, and z and provides a resolution better than 5 nm
[15].
However, there are several challenges while performing
micro/nanomanipulation with most of these technologies.
Notable shortcomings of the instruments based on SPM are lim-
ited number of DOFs and small travel ranges; they are capable
of limited manipulation in the x–y plane with a travel range of
100 ␮m and very small translation (a few micrometers) in the
vertical direction with little rotational capability. There are several limitations with piezoelectric actuators as well. They are
sensitive to environmental changes such as temperature. Moreover, severe nonlinearities are present in the dynamic behavior
of PZT [16]. Well-known phenomena of hysteresis and creep,
along with nonlinear voltage dependence affect negatively the
dynamic response, sometimes precluding closed-loop operation
[17]. Furthermore, the voltage required to operate piezoelectric
actuators can be as high as several-hundred volts.
A novel positioning technology is required to overcome
the limitations of prevailing techniques discussed above and
to fulfill the performance specifications for nanomanipulation
and nanofabrication processes. Previous research has shown
that magnetic levitation is a promising technology for such
applications. Kim and Trumper developed and demonstrated a
high-precision planar maglev stage with large planar motion
capability which had a 10 nm resolution and 100 Hz control
bandwidth [18]. It also eliminated stiction, backlash, and hysteresis that limit the precision in position control. A maglev
scanning stage with a 0.6 nm three-sigma horizontal position
noise was fabricated and demonstrated by Holmes et al. [19].
The development and motion control of a large travel ultraprecision magnetic-suspension stage was presented by Kuo et al.
[20]. Khamesee et al. demonstrated the application of magnetic
levitation in a microrobotic system used for the transportation and assembly of miniature parts [21]. This microrobot
can be remotely operated in 3 DOFs in an enclosed environment by transmitting magnetic energy and optical signals from
outside.
In this paper, we present the design and precision construction of two novel maglev stages for nanoscale positioning. This
paper is organized as follows. Section 2 gives a brief overview
of the two positioners and compares their benefits with other
prevailing precision positioners. Section 3 presents the actuator
design and assembly for the two positioners. Design of other
key mechanical components, viz. the platens and base plates is
discussed in Sections 4 and 5, respectively. Identification of the
transfer functions and control system design are presented in
Section 6. Following is a summary of the performance of the
two positioners and their characterization in the experimental
results section, Section 7.
2. Overview of the maglev stages
In this section, we give a brief overview of the two maglev
positioners. The developed maglev instruments are compared in
terms of their characteristics and performance specifications. A
comparison with other prevailing technologies is also given to
demonstrate the benefits of magnetic levitation, and in particular,
the performance of the positioners we developed. Key applications of these maglev devices include nanoscale positioning and
manipulation of micro-assemblies and manufacturing of their
parts, microstereolithography, vibration isolation for delicate
instrumentation, and microscale rapid prototyping.
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
339
Table 1
Comparison with other technologies
Fig. 1. Photograph of the -stage.
2.1. -stage overview
Fig. 1 shows a photograph of the -maglev stage. “” refers
to the shape of the single-moving levitated platen, which consists
of a triangular aluminum part and six single-axis actuators [22].
The compact maglev stage uses the minimum number of actuators required for 6-DOF motion. Each vertical actuator consists
of a cylindrical magnet and a coil, and each horizontal actuator consists of two cylindrical magnets and a coil. Prevailing
precision positioning devices like STMs and AFMs are able to
position in the travel range of 100 ␮m in 3 DOFs. However,
our -maglev stage is capable of motion control in all 6 DOFs
with the travel ranges of 300 ␮m in the x-, y-, and z-translations
and 3.5 mrad in the x-, y-, and z-rotations. The position resolution is better than 5 nm with position noise less than 2 nm rms
[23–25].
2.2. Y-stage overview
Fig. 2 shows a photograph of the Y-maglev stage. “Y” refers
to the shape of the levitated single-moving platen. The Y-stage
possesses several competitive advantages over the -stage in
terms of travel range, payload capacity, simplicity in mechanical
design, and power consumption:
Technologies
Travel range
(␮m)
Resolution
(nm)
SPM scanner [5]
120 (x, y)
6 (z)
Flexure mechanism [14]
0.1 (x, y, z)
100 (x, y, z)
-Stage
300 (x, y)
300 (z)
2
33
150
9
Y-stage
5000 (x, y)
500 (z)
3
33
1667
15
1.83
10
150
Dynamic
range (×103 )
66
0.01
0.67
1. Larger travel range: This second-generation Y-stage has the
planar travel range more than 15 times wider (5 mm × 5 mm)
than that of the -stage. Due to sensor specifications, this Ystage is currently limited to move 500 ␮m in the z-axis and
0.1◦ in rotation. However, the mechanical design of the stage
makes it capable to rotate about 5◦ and translate about 7 mm
in z with proper sensors.
2. Higher payload capacity: The payload capacity of the Ystage is 2 kg as compared to 0.3 kg of the -stage.
3. Fewer number of parts: The Y-stage is based on a novel
actuation mechanism in which forces in two perpendicular
directions are generated by a single magnet (Refer to Section
3.2 for details). This reduced the number of magnets by a
factor of three compared with that in the -stage.
4. No mechanical restriction: The platen of the Y-stage can
be removed from the stage frame easily without disturbing any stationary parts. This facilitates easy loading and
unloading of objects on the platen for various applications,
which is an added advantage over the nanomanipulator
developed by Culpepper and Anderson [14] and the stage [23].
Table 1 gives a comparison of the developed maglev stages
with two prevailing positioning technologies [5,14] in terms of
the travel range, position resolution, and dynamic range (travel
range divided by position resolution).
3. Actuator units
3.1. Actuator units for the -stage
3.1.1. Actuator design
The force between the current-carrying coil and the permanent magnet is calculated by the Lorentz law
f = (J × B) dV
(1)
Fig. 2. Photograph of the Y-stage.
where J is the current density (A/m2 ) in the coil, B the magnetic flux density (T) generated by the permanent magnet, and
dV the small volume segment in the coil. The magnet assembly (described in the following subsection) produces a magnetic
field on the concentric current-carrying coil. The magnetic
340
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
flux density is related to the magnetic scalar potential Ψ as
B = −μ0 Ψ , where Ψ was obtained by the following superposition integral [26] and is given by
ρm (r )
Ψ=
(2)
dV 4πμ0 |r − r |
In Eq. (2), ρm is the volume magnetic charge density in the
magnet, and r and r are the position vectors to an observer
on the coil and an infinitesimal volume segment dV in the
magnet, respectively. Then, the resulting force can be found
from the triple integral of J × B over the whole coil volume
V . The design calculations were performed using MathCAD.
Because of the difficulty in deriving a closed-form expression,
we chose to calculate the force with respect to a set of given
values of z, and used the “polyfit” command in MATLAB
to obtain a gap-dependent current–force relation using leastsquare polynomial fit. In the operating travel range of 300 ␮m,
however, this force constant essentially remains constant. A
detailed explanation of the design calculations was reported in
Ref. [24].
3.1.2. Assembly
The six-axis motion generation by the platen for the -stage
is achieved by the application of a combination of independent force components acting through six unit actuators. Each
horizontal-actuator unit consists of two cylindrical permanent
magnets attached together with an aluminum spacer between
them placed inside a doughnut-shape current-carrying coil. Each
vertical-actuator unit consists of one permanent magnet and a
current-carrying coil identical to those in a horizontal-actuator
unit. The assembly of the magnets and spacer is attached to the
moving platen, while the six current-carrying coils are fixed to
the stationary base plate.
Schematics of the assemblies of the horizontal and vertical
actuators are shown in Fig. 3. The coil generates the N or S pole
based on the direction of the current governed by the right-hand
rule. Depending on its magnetization direction, an attractive
or repulsive force is applied on the magnet. A magnet-coil air
gap of 0.5 mm limits the total achievable travel range of the
platen.
All the magnets and aluminum spacers/mounts were fixed
using PC-7 epoxy with an 8 h curing time. The precision tooling was designed and constructed for the assembly of these
actuator units. A pocket of a 11.5 mm square with the round
corners of a 3.175 mm radius was machined. The height of the
tooling for the vertical actuators is 13.513 mm, and for horizontal tooling, 22.987 mm. The surfaces to be fixed were cleaned
with methyl ethyl ketone (MEK) and scratched with sandpaper which increased the bonding strength. All the surfaces to
be fixed were applied a thin layer of epoxy. Spray of dry-film
mold release was applied on the tooling so that the magnets and
spacer/mount would not stick to the tooling. Then magnet assemblies were placed in the tooling and were squeezed and held
between two flat surfaces with G-shaped clamps. The assemblies were kept for at least 24 h for curing and were then taken
off. The pocket sides in the tooling of the exact size ensured
a very precise alignment of the magnets and spacer/mount.
The preset height of the tooling ensured the uniform epoxy
thickness and overall length from one magnet assembly to
another.
The primary benefit of this maglev stage apart from its precision positioning and load-carrying capabilities is the nominal
power consumption by the actuators. The average nominal current in each vertical actuator is 0.7 A, so the power consumption
with the coil resistance of 0.6 is only around 1 W in the entire
actuation system to levitate the platen against gravity. Assuming the coils to be point sources of heat, this power consumption
would increase the temperature of the base plate by a maximum
of 0.2 ◦ C. Furthermore, this heat is transferred to the platen
only through convection since there is no mechanical contact
between the base plate and the platen. Thus, for all practical
purposes, the thermal expansion errors due to heat losses in the
positioner that would be detrimental to its nanoscale positioning capabilities may be safely ignored. The characterization of
these actuators was described in detail in Ref. [22]. The detailed
specifications of the actuating units in the -stage are given in
Table 2.
Table 2
Properties of actuators [27]
Permanent magnet
Coil
Fig. 3. Schematics of (a) the horizontal-actuator unit and (b) the vertical-actuator
unit.
Property
-Stage
Y-stage
Size (mm)
Height (mm)
Material
Energy product
(kJ/m3 )
Wire gauge
Material
Turns
Resistance ()
Inner dimension
(mm)
Outer dimension
(mm)
Thickness (mm)
Ø 11.7
9.5
NdFeB
400
25.4 × 25.4
12.7
NdFeB
280
AWG#24
Copper
179
0.6
Ø 12.2
AWG#24
Copper
679V ; 561H
5.5V ; 5.9H
10 × 10V ; 20 × 20H
Ø 32.5
35 × 35V ; 40 × 40H
9.6
17.5
V: vertical actuator, H: horizontal actuator.
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
341
Fig. 4. Cross-sectional side view of the novel two-axis actuator.
3.2. Actuator units for the Y-stage
3.2.1. Actuator design
The working of the Y-stage is based on a novel 2axis electromagnetic-force generation scheme, which generates
forces on a single permanent magnet in two perpendicular directions using two current-carrying horizontal and vertical coils
[27]. The terms “vertical coil” and “horizontal coil” are used for
the coils to generate vertical and horizontal actuation, respectively. Fig. 4 shows a cross-sectional view of each actuator unit.
There are three such units at the three ends of the Y-shaped
platen shown in Fig. 2. The magnetic-field lines generated by
the permanent magnet are also shown in Fig. 4. The directions
of the currents in flow are assumed clockwise in the vertical
coil and counterclockwise in the horizontal coil, seen from the
top. The maglev stage is comprised of three such actuating units
with three permanent magnets attached to ends of the Y-shaped
platen and two fixed square-shaped coils per magnet. The magnetic flux generated by the magnet is shared by the two coils, one
right below and another on the side of the magnet. The magnitude
and directions of currents in the coils govern the forces exerted
on the magnet following the Lorentz-force law. In the vertical
coil, the direction of the magnetic-field lines is normal to the
direction of current flow and towards the center of the coil on all
the four sides of the coil (neglecting the corner effect). Thus, the
Lorentz force on the coil is vertically downwards on all the four
sides of the square. The equal electromagnetic reaction force is
applied vertically upwards on the moving magnet since the coil
is fixed in a stationary frame. Similarly, in the horizontal coil
the direction of the magnetic-field lines is approximately downwards in all the four sides. Thus, the effective force on the coil is
to the right following the right-hand rule, and on the magnet, to
the left. To change the magnitude and directions of the vertical
and horizontal forces, we make appropriate changes in the magnitude and directions of current flow in the corresponding coils.
In this manner we can generate the forces in the two perpendicular directions independently on a single-moving magnet. The
detailed specifications of the actuating units in the Y-stage are
given in Table 2.
The force generated by the interaction between the permanent
magnet and the coil current follows the Lorentz force given by
Eq. (1). The limits of the integral are set to cover the entire
volume of the coil. After substituting the physical parameters of
the magnet and the coil, the expression of the force acting on the
volume (V) of the coil due to the surface (S) magnetic charge on
Fig. 5. Exploded views of (a) the vertical-coil and (b) the horizontal-coil assemblies.
the magnet becomes the following quintuple integration.
f =
Jσm
4π
S
⎛
×⎝
V
⎞
−(x − p)k̂+(z − r1 )î
(x − p)2 + (y − q)2 +(z− r1 )2
3/2 dy dx dz⎠ dp dq
(3)
In the above equation, σ m = ±μ0 M is the surface magnetic
charge density on the top and bottom surfaces of the magnet
with permanent magnetization M (A/m). The permeability of
free space μ0 is 4π × 10−7 H/m, and îand k̂ are the unit vectors in the stationary frame attached to the coil; p, q, and r1
are the integration variables for distance of the surface element
on the magnet from the coordinate origin in the x-, y-, and zaxis, respectively. For the sake of computational simplicity, we
made use of the coil symmetry by dividing it in four sections
along the diagonals and performed the force calculations with the
quintuple integral only on one of them. Details of these design
calculations for optimal actuator sizing, its performance at offset
points, and experimental verification of the final design values
may be found in Ref. [28].
This new actuation scheme has several advantages over the
previous one. (1) It makes the mechanical design of the maglev
stage very simple to manufacture and assemble, (2) there is no
mechanical constraint on the platen to separate it from the actuator assembly, and (3) since there are only three magnets used
to generate actuation forces in all the six axes, the structure of
the platen is simple. The coils need to be attached to the base
plate at precise locations and orientations. The surface of the
coils should also match perfectly with the top plane of the base
plate. Therefore, we designed and fabricated coil holders shown
in Fig. 5 for both the vertical and horizontal coils and fixed them
together using assembly fixtures.
3.2.2. Vertical-coil assembly
Attaching the vertical coil to the base plate was somewhat
cumbersome because this coil could not be held from the outside
and no part of the coil holder can come out of the coil due to its
342
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
one used for the vertical-coil assembly was constructed. The
assembly procedure is shown in Fig. 7. The coil and the coil
holder were bonded in place with PC-7 epoxy, and the clamping plate was coated with dry-film mold release. Then the coil
holder was mounted on the reference plate, and the clamping
plate was fastened after the coil was placed.
4. Platen design and assembly
Fig. 6. Fixture to pot the vertical coil to its coil holder.
tight packaging. Thus, we designed the coil holder as an insert
in the coil and fixed it inside the coil. An exploded view of the
vertical-coil assembly is shown in Fig. 5(a). The coil holder was
fabricated to perfectly fit to the inner surface of the vertical coil
leaving space for epoxy. It has 2 through holes for the screws so
that the coil can be placed at a precise position and orientation.
To bond the vertical coil accurately within its holder we designed
the assembly fixture shown in Fig. 6. The inner surface of the
coil and the outer surface of the coil holder were bonded together
with PC-7 epoxy. The reference plate and the clamping plate
were coated with dry-film mold release thoroughly. The coil
holder was mounted on the reference plate, and then the coil
was placed on the top. A clamping plate was used to press the
coil from the top. This assembly was kept for 24 h to cure the
epoxy.
Fig. 8 shows exploded views of the two maglev positioners.
The triangular part in the center in Fig. 8(a) and the Y-shaped
part in Fig. 8(b) are the single-moving platens. The platens carry
permanent magnets for actuation and plane mirrors for horizontal motion sensing. The details of the platen design and precision
assembly are described in the following sections.
Effective packaging is important for any assembly. Thus, we
designed all parts and their assemblies in SolidWorks® in such
a way that (1) the platens must be close enough to all coils for
sufficient force generation, (2) the bottom surfaces of the platens
must be in the sensing range of all capacitance probes, and (3)
the HeNe laser beams from the interferometers should not be
blocked.
3.2.3. Horizontal-coil assembly
The horizontal coils needed to be placed at the same horizontal level as that of the platen. An inverted-T-shaped coil holder
was designed and fabricated that could hold the horizontal coil
from inside and support it from the bottom. An exploded view
of the assembly is shown in Fig. 5(b). A fixture similar to the
Fig. 7. Fixture to pot the horizontal coil within its coil holder.
Fig. 8. Exploded views of (a) the -stage and (b) the Y-stage.
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
Since the platen is the only moving part in each of the whole
maglev stages, its design is very crucial with the following considerations: (1) the vertical actuators must continuously apply
vertical forces to balance the platen’s weight against gravity at a
predetermined levitation height, (2) the platen should be light to
maintain the coil currents in the actuators low, and (3) the moving and stationary structures should be stiff enough to exhibit a
high natural frequency.
4.1. -stage platen
The core of the platen is made of pocket-milled singlepiece aluminum to reduce its mass and to keep its natural
frequency high without sacrificing stiffness. A finite-element
analysis shows the natural frequency of the platen at as high as
4.6 kHz. There is no iron part in the assembly, which makes the
actuators operable at high bandwidth without magnetic saturation or hysteresis. The parts attached to it are three horizontal
magnet assemblies, three vertical magnet assemblies, three plane
mirrors, and a set of aluminum, viscoelastic, and constraint layers on top. The assembly of these three layers adds passive
damping to the system to minimize the structural vibrations and
improve the stability. The bottommost aluminum layer, called
the top plate, covers the pockets of the platen. It was attached to
the platen using four screws, three at the corners and the fourth
at the center. The total mass of the moving part is 0.212 kg.
The horizontal and vertical actuator assemblies were
described in the previous section. There are three pairs of arms
protruded on the sides of the platen to hold the magnet assemblies for horizontal actuation. Each arm has a hole with its
diameter 10 ␮m larger than that of the cylindrical magnet. The
three setscrews on the outer, top and bottom surfaces of the
arm were tightened to hold the magnet. Since the magnetization
directions of the magnets are parallel to the sides of the triangular
platen, these units generate horizontal force parallel to the sides
of the platen. Each of the three vertical actuator assemblies consists of a cylindrical magnet and a cylindrical aluminum mount
with three holes potted together. Each assembly is mounted on
the platen using three screws.
Fig. 9. Tooling used to fix the plane mirrors to the -platen core.
343
Three plane mirrors for horizontal motion sensing were
mounted on the platen using double-sided tape. These plane
mirrors act as the reflectors for the laser interferometers. Two
of the mirrors are on one side of the platen, and the third, on
the second side at an angle of 120◦ . To ensure satisfactory parallelism of these mirrors a tooling was designed and fabricated.
This tooling and the mirror assembly are shown in Fig. 9. The
bottom surface of the platen was machined and ground to be flat
within 2.54 ␮m as this surface would be used as the targets by
the capacitance probes to sense the vertical displacements of the
platen.
4.2. Y-stage platen
4.2.1. Platen core
The platen core was machined from single-piece aircraftgrade 7075 aluminum. It was pocket-milled in the center leaving
the ribs on the edge to reduce the weight while keeping the high
stiffness and natural frequency. The shape of the pocket was
decided to make its fabrication easy. The width of the pocket,
15.875 mm, can be machined with a single pass of a 5/8 in.diameter end-milling tool. Several designs of the platen core with
different pocket depths were made and analyzed for the mass
and natural frequency. The depth of the pocket was finalized to
be 7.62 mm. The three ends of the Y-shaped platen core have
a width of 25.4 mm and a height of 12.7 mm to match the side
dimension of the magnet. A small hole was drilled and tapped in
the center of platen to tighten a screw to hold a grounding wire
for the capacitance probes. The bottom surface of the platen was
ground to achieve a surface roughness of 2.54 ␮m as this surface
would be used as the targets by the capacitance probes.
4.2.2. Assembly
The assembly process of the platen is shown in Fig. 10. The
three cuboid-shaped (25.4 mm × 25.4 mm × 12.7 mm) permanent magnets were attached around the platen core, and then
Fig. 10. Assembly of the Y-platen.
344
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
Fig. 11. Fixing process of the magnet with the Y-platen core.
three mirrors were attached at one of the side faces of the magnets. The 25.4 mm square faces on the top and bottom are the N
and S poles of the magnets. The faces of the mirrors away from
the magnets are the reflecting surfaces.
The fixture shown in Fig. 11 was fabricated and used for
precision in the gluing process. Several coatings of dry-film mold
release were applied on the fixture so that the magnet or the
platen core does not get stuck to the fixture. The magnet face to
be bonded and the end of platen core were cleaned with MEK
to remove oil and dirt and roughened using sandpaper. A thin
layer of PC-7 epoxy was applied on both the surfaces, and the
magnet was placed in the fixture. The platen core was inserted
from the side and pressed into place. Then the fixture top was
clamped with four screws. This assembly was kept for 24 h to
provide sufficient time for the epoxy to cure. Then the clamping
screws were removed and the assembly was carefully taken off
the fixture, and the same procedure was repeated for the two
remaining ends.
5. Base plate and stationary assembly
5.1. -stage assembly
The base plate has the dimensions of 203.2 mm ×
203.2 mm × 56.0 mm. Fig. 12 shows the bottom view of the
platen showing the pocket that is milled from the bottom surface
of the platen to fasten the screws. All the parts and assemblies
were mounted on the base plate via through holes, tapped holes,
and pockets. Three screws were used to mount each of the three
capacitance probes and fastened from the bottom. There are three
circular pockets machined on the top surface of the base plate.
The purpose of these pockets was to maintain the heights of
the vertical coils at which the vertical magnets will exert the
maximum forces when the bottom surface of the platen is in
the sensing range of the capacitance probes. Since the radial
clearance between the magnets and the coils was just 500 ␮m,
accurate positioning of the coils was crucial. Thus, a fixture was
designed and fabricated to accurately position the coils at the
desired locations while we fastened them to the base plate. This
fixture had three cylinders mounted on a flat plate as shown in
Fig. 13. The diameter of the cylinder was the same as the inner
Fig. 12. Photograph of the bottom view of the base plate with nine M3 holes to
mount the three capacitance probes. Three additional 10–32 brass screws were
inserted to fasten the vertical coils.
diameter of the vertical coils so these cylinders were press fit
to the vertical coils. While assembling, the three vertical coils
were attached to the fixture and placed on the base plate. Then
the three brass screws were tightened from the pocket at the bottom of the base plate. Then, this fixture was slowly taken off.
This ensured the position of the coils to be as close as possible
to the designed.
The coils for the horizontal actuators were fixed to their coil
holders. These coil holders were mounted on the base plate using
four screws. Each holder has four through holes, and there are
tapped holes on the base plate. Precise alignment of the coils was
important while potting because of a low clearance of 504 ␮m
between the outer diameter of the magnet and inner diameter of
the coil, and possible loss of travel range in case of any shift or
rotation of the coil. Thus, a fixture was designed and fabricated
Fig. 13. Tooling for mounting vertical coils to the base plate.
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
345
Fig. 15. The base plate for the Y-stage showing the locations of various holes
to fix the stationary parts.
Fig. 14. Tooling to fix the horizontal coil to its coil holder.
for precision assembly. Fig. 14 shows the fixture and the coil
holder. The coil holder was mounted on the fixture using the
four screws. A cylindrical insert with the diameter the same as
the coil inner diameter was used to hold the coil in position. The
collar on one side and a supporting disc with a bolt on the other
side kept the coil fixed at the desired position relative to the coil
holder. A thin film of epoxy was applied on the surfaces of the
coil holder and the coil. The coil was held at position for 24 h
to cure the epoxy. This process ensured a precise alignment of
the coils with respect to the platen with a tolerance of 76 ␮m (3
mils) by our conservative estimation.
5.2. Y-stage assembly
All the stationary parts are mounted on an aluminum base
plate. The three cylindrical pieces placed in the center are capacitance probes. These capacitance sensors were mounted right
below the platen at 120◦ -apart symmetrical positions and measure the height of the platen at three different locations. The
vertical-axis motions of the platen are sensed by triangulation
of the data gathered from these three capacitance sensors. The
vertical-actuator coils were placed adjacent to the capacitance
probes. Next to the vertical coils are the horizontal coils mounted
with the horizontal-coil holders. Fixed at the edges of the base
plate are three terminal strips. Each terminal strip has four contacts to make connections to one vertical and one horizontal
coil.
Fig. 15 shows the base plate that holds all the coil assemblies, capacitances probes, and terminal strips. The base plate
has an area of 200 mm × 200 mm and a height of 44 mm. The
area was designed so that the base plate could hold all the stationary components of the vertical and horizontal actuators and three
capacitances gauges. The height was chosen so as to align the
plane mirrors on the platen for horizontal motion sensing with
the laser beam from the laser head. It was fabricated from solid
aluminum to keep its thermal capacity high so that repeated variations in current would not lead to significant thermal-expansion
errors. Specific locations of holes for fixing the stationary parts
of the assembly are shown in Fig. 15.
Table 3
Summary of maglev-stage specifications
Property
Moving elements
DOFs
Mass (kg)
Moment of inertia (10−6 kg m2 )
Size (mm)
Y-stage
-stage
6
0.212
132.88 −3.14
−3.14 122.28
0
0
80 × 69 × 17.5
6
0
0
235.87
0.267
340.37
0
0
340.37
0
0
115 × 127 × 12.7
Actuation
Number of coils
Number of magnets
Power consumption (W)
6
9
1.0
6
3
0.8
Motion capabilities
Travel range (trans.) (mm)
Travel range (rot.) (mrad)
Max acceleration (ms−2 )
Max payload (kg)
Position resolution (nm)
Angular resolution (␮rad)
0.3 × 0.3 × 0.5
3.5 × 3.5 × 3.5
30
0.3
5
300
5 × 5 × 0.5
3.5 × 3.5 × 3.5
80
2.0
4
100
0
0
653.61
346
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
6. Control
6.1. Controller design
In the absence of any restraining force, the independent
plant transfer functions (TFs) for the translational and rotational
motions may be respectively represented as
X(s)
1
=
F (s)
ms2
(4)
Θ(s)
1
= 2
T (s)
Is
(5)
The values of mass m and moments of inertia I for the two
positioners are reported in Table 3.
We used the MATLAB function SISOTOOL to select the control parameters. The design method was the same for controllers
in all six axes for both the positioners. For the -stage, we set the
damping ratio ζ to be 0.7 and the phase margin, 52◦ at a crossover
frequency of 50 Hz. The lead-lag compensator designed for the
z-axis motion control for this stage is given by Eq. (6).
Gz (s) =
8.7315 × 104 (s + 135)(s + 11)
s(s + 1385)
Fig. 17. Closed-loop frequency responses of the analytical model (dashed line)
and the plant TFs from the FFTs of the input-output signals (solid line).
The value of K for z is 2.32003 × 105 N/m and 2.95658 × 102 N
for θ and ψ.
All the designed controllers were converted to discrete-time
difference equations using the zero-order-hold (ZOH) equivalence method with a 5 kHz sampling rate. The control laws were
implemented in C on a digital signal processor with the user
interface running on VME PC for the real-time control of the
positioners.
(6)
The free pole at the origin in the s-plane is to eliminate the
steady-state error. We used the same controller for the x- and yaxis motion control. The controllers for the three rotational axes
have the same pole-zero locations with the controller gains of
96.872 N, 54.575 N, and 50.221 N for φ, ψ and θ, respectively.
For horizontal motion control of the Y-stage, a continuoustime lead-lag controller was designed with phase margin of 69.7◦
at the crossover frequency of 109.5 Hz. The designed compensator is given by
Gx,y,φ (s) =
K(s + 116.6)(s + 10.91)
.
s(s + 4014)
(7)
where the controller gain K for x and y is 7.3726 × 105 N/m and
1.8041 × 103 N for φ.
The design of three SISO controllers for vertical motion
control was performed similarly. A continuous-time lead-lag
controller was designed to control the vertical motions that has
phase margin of 70.1◦ at crossover frequency of 65 Hz. The
compensator is given as
Gz,ψ,θ (s) =
K(s + 57.47)(s + 6.271)
s(s + 2103)
Fig. 16. Block diagram for stochastic transfer function identification.
(8)
Fig. 18. Position noise in (a) the -stage and (b) the Y-stage.
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
6.2. Experimental plant TF identification
System identification is necessary in order to verify the analytically determined plant TFs. It is also necessary to analyze and
model the plant behavior accurately, to subsequently develop
effective controllers. In case of the two maglev positioners
reported herein, system identification is particularly challenging
because of their open-loop instability. Thus, the system identification needs to be performed in closed loop after the platen
is stabilized around the operating point with decoupled lead-lag
controllers described in the previous subsection.
Since we have full control over the excitation signal, it is
desired to use the signals that persistently excite the plant. Thus,
the model validation was performed with a zero-mean, whiteGaussian noise (WGN) as the reference input signal. A detailed
treatment of stochastic modeling is covered in classical texts
such as [29]. Fig. 16 shows the block diagram to obtain the plant
TF. The random disturbance is generated in the software with
Fig. 19. (a) A 10 nm step response of the -stage and (b) 50 nm step response
of the Y-stage.
347
the “rand ()” function in the C language. The magnitude of the
random input signal was increased gradually from 0.1 to 10 ␮m.
These numbers were chosen to excite the plant persistently without loosing stability. The total time of the excitation was 2 s with
a sampling rate of 5 kHz.
Fig. 20. (a) A 15 nm and (b) 15 ␮m consecutive steps generated by the Y-stage
with command-tracking errors.
348
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
The experimental result of the system identification using
the input-output time sequences is presented in Fig. 17. The
figure shows the analytical closed-loop system response of the
Y-stage in the x-axis with the controller given by Eq. (7), and
the identified TF. System identification in other five axes was
performed in a similar manner. From the figure, it may be noted
that the analytical and experimental plant TFs match closely.
Therefore, for the rest of the experimental analysis, we have
used the analytical plant models Eqs. (4) and (5) for controller
design and performance testing.
7. Experimental results
The characteristics and performance specifications of the
two maglev stages are summarized in Tables 1 and 3. Sev-
eral experiments were conducted to illustrate the motion-control
capability of the maglev stages in terms of position regulation,
step responses, and multi-axis contouring. These results have
been reported in detail in Ref. [27]. Here we present some of
these experimental results to validate the values reported in the
tables. These results were taken in the x-axis motion of both the
stages as the performance of the y-axis is not much different
from that of the x-axis.
The experimental results on the position noise for horizontal
motion of the -stage is shown in Fig. 18(a). The position regulation is around 2 nm rms. Fig. 18(b) shows a position noise
profile for horizontal motion of the Y-stage, which is better than
3 nm rms. The position noise for the vertical motion is higher due
to sensor noise, analog-to-digital conversion (ADC) noise, and
ADC quantization. Despite the noise in the vertical axes the hor-
Fig. 21. (a) A 50 nm-radius circle traversed by the -stage, (b) 50 nm-radius circle traversed by the Y-stage, and (c) 2.5 mm-radius circle traversed by the Y-stage.
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
Fig. 22. A 5 mm ramp motion generated by the Y-stage.
izontal motion profiles are very quiet. This indicates that there
is little dynamic coupling between the horizontal and vertical
motions of the maglev stages.
Fig. 19(a) and (b) respectively shows a 10-nm step response
of the -stage and the Y-stage. The position resolution is clearly
better than 5 nm and 4 nm, respectively. Fig. 20 shows experimental results of consecutive steps with the Y-stage. The step
sizes are 15 nm and 15 ␮m, respectively. The error plots show
that even with the scaling by a factor of 1000, the commandtracking errors were within ±10 nm in both cases.
In the real-time control code written in the C language as an
interrupt service routine (ISR), a set of data points in more than
one axis can be allocated with respect to time so that the platen
can follow a multi-axis trajectory. In Fig. 21(a), a commanded
path of a 50 nm-radius circle and the response of the -stage
are shown. Fig. 21(b) shows a 50 nm circle traced by the Yplaten. Fig. 21(c) shows a 2.5 mm-radius circular path traversed
by the Y-stage in the x–y plane. The motions of the platens were
very close to the perfect circles with command-tracking errors
within ±10 nm. These test results demonstrate that the maglev
devices can track small multi-axis trajectories and achieve position control at nanoscale precision. It also demonstrates that for
the Y-stage with a larger travel range, the positioning capabilities
are equally good at micro as well as nanoscale.
The large travel range of the Y-stage has also been demonstrated by a ramp response of 5 mm in the x-axis as shown in
Fig. 22. The platen covered the entire travel range of 5 mm from
end to end.
8. Conclusions
Precision in nanomanipulation has become the key to the
development of the next-generation of nanotechnology. The
demands of ultra-high-precision positioning stages are continuously increasing. Currently available stages are able to provide
good resolution but over very short travel ranges on the order
of several 100 ␮m. This limitation will significantly affect the
nanoassembly processes when the process needs to be completed
349
by several different tools. Moreover, prevailing stages have no or
very small vertical translation capabilities, which limits their use
in micromanufacturing processes. The current nanotechnology
requires nanomanipulation stages that can move and position the
specimen to various locations and in desired orientations. This
motion needs to be in multiple DOFs with long travel ranges at
nanoscale resolution.
This paper presented two novel maglev stages that fulfill all the requirements mentioned above. Particular attention
was given to the design and precision construction of the
actuators and other mechanical components. Additionally, the
control system design and experimental results demonstrating
the nanopositioning capabilities of the two positioners were
also presented. These maglev stages have very simple mechanical structures with a single-moving part, which is easy and
inexpensive to fabricate. The constituting components have no
mechanical contact between the moving and stationary parts,
which facilitates the maintenance of high position resolution.
This also eliminates wear in the mechanical parts thereby
increasing their life spans and eliminating the requirement of
lubricants. Thus, the maglev stages presented in this paper can
be used in the nanomanipulation and nanomanufacturing in the
next-generation nanotechnology.
Acknowledgment
This material is based on the work supported by the National
Science foundation under the Grant No. CMS-0116642.
References
[1] U.S. Department of Energy. Nanoscale science, engineering and technology
research directions.
[2] Sitti M. Survey of nanomanipulation systems. In: Proceedings of the IEEE
conference on nanotechnology. 2001. p. 75–80.
[3] Bauer C, Bugacov A, Koel BE, Madhukar A, Montoya N, Ramachandran
TR, et al. Nanoparticle manipulation by mechanical pushing: underlying
phenomena and real-time monitoring. Nanotechnology 1998;9(4):360–4.
[4] Yu MF, Dyer MJ, Rohrs HW, Lu XK, Ausman KD, Her JV, et al. Threedimensional manipulation of carbon nanotubes under a scanning electron
microscope. Nanotechnology 1999;10(3):244–52.
[5] Veeco
Instruments.
Applications
of
AFM.
http://web.mit.edu/cortiz/www/nanomechanics.html.
[6] Zhang B, Zhu Z. Developing a linear piezomotor with nanometer resolution
and high stiffness. IEEE/ASME Trans Mechatron 1997;2(1):22–9.
[7] Dong L, Arai F, Fukuda T. 3D nanorobotic manipulation of nano-order
objects inside SEM. In: Proceedings of 2000 international symposium on
micromechatronics and human science. 2000. p. 151–6.
[8] Carter FM, Galburt DN, Roux S. Magnetically levitated and driven reticlemasking blade stage mechanism having six degrees freedom of motion. US
Patent 6,906,789; 2005.
[9] Hollis RL, Musits BL. Electromagnetic X-Y-Theta precision positioner.
US Patent 4,514,674; 1985.
[10] Galburt D. Electro-magnetic apparatus. US Patent 4,506,204; 1985.
[11] Sun Y, Piyabongkarn D, Sezen A, Nelson BJ, Rajamani R, Schoch R, et al.
A novel dual-axis electrostatic microactuation system for macromanipulation. In: Proceedings of IEEE/RSJ international conference on intelligent
robots and systems. 2002. p. 1796–801.
[12] Mori S, Hoshino T, Obinata G, Ouchi K. Linear actuator with air bearing
for highly precise tracking [HDD]. In: Digest of the Asia-Pacific magnetic
recording conference. 2002. p. AP4-01-2.
350
W.-j. Kim et al. / Precision Engineering 31 (2007) 337–350
[13] Egashira Y, Kosaka K, Takada S, Iwabuchi T, Baba T, Moriyama S, et al.
0.69 nm resolution ultrasonic motor for large stroke precision stage. In:
Proceedings of IEEE-NANO. 2001. p. 397–402.
[14] Culpepper ML, Anderson G. Design of a low-cost nanomanipulator
which utilizes a monolithic, spatial compliant mechanism. Precis Eng
2004;28(4):469–82.
[15] Zyvex Corporation. Zyvex F100 specifications and picture.
http://www.zyvex.com/Products/F100 Features.html#.
[16] Richter H, Misawa E, Lucca D, Lu H. Modeling nonlinear behavior in a
piezoelectric actuator. Precis Eng 2001;25(2):128–37.
[17] Requicha AAG, Meltzer S, Terán AFP, Makaliwe JH, Sikén H, Hsieh S,
et al. Manipulation of nanoscale components with the AFM: principles
and applications. In: Proceedings of IEEE conference on Nanotechnology.
2001. p. 81–6.
[18] Kim WJ, Trumper DL. High-precision magnetic levitation stage for photolithography. Precis Eng 1998;22(2):66–77.
[19] Holmes M, Hocken R, Trumper DL. The long-range scanning stage: a novel
platform for scanned-probe microscopy. Precis Eng 2000;24(3):191–209.
[20] Kuo SK, Shan X, Menq CH. Large travel ultra precision x-y-θ motion
control of a magnetic-suspension stage. IEEE/ASME Trans Mechatron
2003;8(3):334–41.
[21] Khamesee MB, Kato N, Nomura Y, Nakamura T. Design and control
of a microrobotic system using magnetic levitation. IEEE/ASME Trans
Mechatron 2002;7(1):1–13.
[22] Kim WJ, Maheshwari H. High-precision control of a maglev linear actuator with nanopositioning capability. In: Proceedings of American control
conference. 2002. p. 4279–84.
[23] Verma S, Kim WJ, Gu J. 6-axis nanopositioning device with precision magnetic levitation technology. IEEE/ASME Trans Mechatron
2004;9(2):384–91.
[24] Gu J, Kim WJ, Verma S. Nanoscale motion control with a compact
minimum-actuator magnetic levitator. ASME J Dyn Syst Meas Control
2005;127(3):433–42.
[25] Verma S, Kim WJ, Shakir H. Multi-axis maglev nanopositioner for precision manufacturing and manipulation applications. IEEE Trans Ind Appl
2005;41(5):1159–67.
[26] Haus HA, Melcher JR. Electromagnetic fields and energy. Prentice-Hall,
Inc.; 1989.
[27] Verma S. Development of novel high-performance six-axis magnetically levitated instruments for nanoscale applications. PhD Dissertation.
College Station, TX: Dept. of Mech. Eng., Texas A&M University;
2005.
[28] Verma S, Shakir H, Kim WJ. Novel electromagnetic actuation scheme
for multi-axis nanopositioning. IEEE Trans Magnetics 2006;42(8):2052–
62.
[29] Ljung L, Soderstorm T. Theory and practice of recursive identification.
MIT Press; 1983.